Dense weather data made readable, for people and for their agents.
Skylar blends many forecast models, checks them against real stations, and tells you what it means for your day in plain sentences, with a clear verdict and the chart to back it up.
How it works
A forecast starts as millions of numbers from models that rarely agree. Skylar does the reading, so you get the part that matters.
Blends a dozen+ weather models (HRRR, RRFS, NBM, GFS, ECMWF and more) into one consensus.
Corrects model bias with real observations from thousands of stations. Readings are quality-checked first, and each correction fades the further ahead the forecast looks.
What to expect, whether it suits your plans, and when. Written the way a forecaster would say it to a friend.
Updated continuously; the most detailed data runs in 15-minute steps for the next hours.
| 2 pm | HRRR | NBM | GFS | ECMWF |
|---|---|---|---|---|
| TMP K | 291.4 | 290.8 | 292.6 | 291.1 |
| DPT K | 276.2 | 275.9 | 277.8 | 276.4 |
| GUST m/s | 11.8 | 9.4 | 13.1 | 10.2 |
| APCP 1h | 0.00 | 0.00 | 0.25 | 0.00 |
| APCP 4 pm | 0.38 | 0.21 | 0.94 | 0.30 |
| VIS m | 24135 | 16093 | 24135 | 20000 |
| REFC dBZ | 31.5 | – | – | – |
Denver, one afternoon. Kelvin, metres per second, kilograms per square metre. Four models, four answers.
Denver this afternoon
Skylar forecast
Mild and dry through lunch. A few showers possible after 3 pm, clearing by evening. Breezy at times, gusts near 25 mph.
The app
Every screen and card below comes from the current build of the app.
800+ activities, from hiking and golf to running, fishing and pouring concrete. Each one has its own limits, so Skylar can give a straight verdict and point to the best hours.
Ask the way you’d say it: tomorrow morning
, this weekend
, after work next week
. Skylar finds the window.
Ask a weather question and get an answer you can read in seconds, with the chart that explains it.
Pick the things you do. Each favorite gets a home card with its week: the best day, the best hours and a grade for every day.
Track a plan or a trip, like a hike on Saturday or a drive on Sunday. Skylar keeps checking the forecast and tells you when it changes.
The weather at each stretch of road, at the time you’ll be there. Skylar grades the trip and can look for a better time to leave.
A radar loop and maps drawn in Skylar’s own style, with official warnings when they matter.
The detail is there when you want it, a tap deeper.
How Skylar writes
A bare number leaves the forecasting to you. Skylar says what it means, in the words a forecaster would use. The numbers are still there, one tap deeper, for when you want them.
Instead of an hourly rain probability
A few showers possible late, clearing by midnight.
Instead of a gust table
Breezy after lunch, with gusts near 30 mph on the ridges.
Instead of a dewpoint and a low
Frost likely by dawn. Cover the tomatoes.
Instead of a spread of model runs
Still uncertain this far out. Firming up by Thursday.
Coming soon · Preview
Skylar is planned as an MCP server, tied to Skylar accounts, with a plain REST twin for agents that don’t speak MCP. An assistant will be able to ask for forecasts, activity verdicts, best windows and route weather, and get back the same readable answers people see.
Later, with a linked account: schedule_activity, watch_forecast, list_favorites
Preview only. Names and shapes may change before launch.
Answers come back as compact words and numbers an agent can quote, not raw model output.
Person
Is Saturday morning good for a hike near Boulder?
Agent calls Skylar
evaluate_activity { "activity": "hiking", "place": "Boulder, CO", "when": "Saturday morning" }
Skylar returns
Agent replies
Saturday morning looks good. Go early: it’s dry from 7 to 11 with light wind, around 52° at the start and 68° by late morning.
Agent calls Skylar
get_forecast {"place": "Denver, CO"}
Skylar returns
Illustrative. Real responses will follow the published schema.
Plans
The everyday forecast, done well.
For people who plan around the weather.
For weather-heavy work and storm season.
Prices will be announced at launch.
